Meta's Muse Ember costs $0.10 per million input tokens and targets the repetitive work agents do all day

Meta just priced a model for the boring 90% of agent work, and that is where most business automation lives.

What happened

Meta Superintelligence Lab released Muse Ember on November 9, its lightest and most cost efficient model. Meta says it is designed for "the work agents repeat all day: tool calls, coding subtasks, routing, extraction, and long document parsing," TechCrunch reports.

It costs $0.10 per million input tokens and $0.50 per million output tokens through the Meta Model API and Muse Code, according to The Decoder. It handles text, images, audio and video and takes up to one million tokens at once. Meta says it delivers more intelligence per dollar than Claude Haiku 5.5 or GPT-6 Luna, and in the Artificial Analysis test it cost about $0.14 per task for a score of 36, versus $2.39 for Muse Spark 1.3 at 37. Not every number favors Ember: Haiku scored 34.9% to Ember's 27.2% on FrontierCode 2.0-AA coding tasks.

Meta pitches Ember as the worker under a bigger model. In its words, "Muse Wonder plans, Muse Ember handles the volume." TechCrunch's example is a support agent where a heavy model sets the rules and a light one applies them across tens of thousands of cases.

My take

Extraction, routing and parsing long documents are exactly the steps that eat token budgets in real client workflows: invoices, intake forms, call recordings, scanned contracts. A capable model at this price changes which jobs are worth automating at all.

Before swapping it in:

1. Test on your own documents. Benchmark scores are averages, and your client's messy PDFs are not.
2. Keep the model behind one switch in your workflow so you can change providers without a rebuild.
3. Track accuracy per field, not per document, so one bad field does not hide in a passing run.

And note the coding gap. For steps that write or fix code, the cheaper option may not be the cheaper result.